errbase

• 

  Errbase CLI

Inspiration

Every developer has been there. You finally fix a frustrating terminal error, move on with your work, and a few weeks later the exact same error appears. Instead of remembering the solution, you end up searching Google or Stack Overflow all over again. We wanted to build something that could remember those fixes for you, so your terminal learns from your past instead of making you repeat it.

What it does

Errbase remembers the errors you have already solved. Whenever the same or a similar error shows up again, it suggests the fix that worked before. Instead of storing everything as plain command history, it uses a knowledge graph to connect errors, fixes, tools, and projects. This makes finding the right solution much faster and more reliable.

How we built it

We built Errbase using Python and FastAPI for the backend, Neo4j as the knowledge graph, and React for the frontend. Gemini helps us understand and compare error messages so similar issues can be matched even when they are not exactly the same. When an error is solved, Errbase stores the relationship between the error and its fix, making it available for future use.

Challenges we ran into

One of the biggest challenges was deciding when an error had actually been fixed. We also had to deal with noisy terminal output and make sure similar errors could be recognized even if the messages were slightly different. Designing a graph that stays efficient as more data is added was another interesting challenge.

Accomplishments that we're proud of

We are proud that Errbase actually learns from previous fixes instead of just storing terminal history. Seeing an old error appear and having the solution instantly suggested felt like the project was solving a real problem. We also built a clean foundation that can continue improving as more errors are collected.

What we learned

This project taught us that developer knowledge is connected, and a knowledge graph is a much better way to represent it than a simple log file. We also learned how combining graph databases with LLMs can create a practical tool that saves developers time during debugging.

What's next for errbase

We want to make Errbase available as a VS Code extension and support more shells and programming languages. We also plan to let teams share their knowledge graphs so everyone can benefit from previously solved issues. In the future, we want Errbase to explain why a fix works, suggest solutions with confidence scores, and become a debugging assistant that grows smarter over time.

Built With

• cognee
• fastapi
• python
• rich